Predicting Depression from Socio-Economical Factors

With the sudden arrival of the Covid19, people have been experiencing higher levels of pressure which has given rise to the number of cases of Depression, a mental illness which can further lead to suicide and contribute to global burden, and therefore, faster ways for its diagnosis need to be sought to prevent further fatalities. Machine Learning, which has been implemented with proven results across various sectors especially for the prediction of diseases, promises to be of great help to facilitate the detection of depression within patients. However, up to now, most implementation for mental health prediction are greatly built on clinical data which takes time to be generated as several tests need to be taken which is not always plausible for everyone. Subsequently, this work explores the application and evaluation of three different machine learning algorithms; ANN, CART DT and SVM, for the prompt prediction of depression based on readily available data; behavioural and social-economical characteristics, and thus provides the best algorithm which could be applied in the context. SVM produced the highest accuracy and was therefore tuned whereby the highest recorded accuracy was of 93.85%. As this work has shown promising results, it is further recommended that future works explore deeper into assessing depression severity.